Skip to main content

Table 7 Model performance on each corpus, measured by micro-F1

From: JCBIE: a joint continual learning neural network for biomedical information extraction

Corpus

NER

SP

ET

\({\text {ET}}^+\)

RE

\({\text {RE}}^+\)

\({\text {ADE}}_1\)

86.58

88.40

87.16

99.82

72.14

91.75

\({\text {ADE}}_2\)

90.66

92.18

90.72

99.81

83.37

98.64

DDI

95.48

96.74

96.42

100

80.00

84.26

CPR

88.58

90.59

89.19

97.97

65.36

74.27

Avg.

90.32

91.98

90.87

99.40

75.22

87.23

  1. NB: NER means SP and ET labels are combined as a single label. ET and ET\(^{+}\) denote the ET predictions depending on SP-predicted labels and gold labels, respectively. RE and RE\(^{+}\) denote the RE predictions depending on SP and ET predicted labels and gold labels, respectively
Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us